Performance analysis and optimization of staggered fin heat exchangers under varying altitudes using machine learning

In this study, we conduct a comprehensive investigation into the flow and heat transfer behaviors of staggered fin heat dissipation channels across varying altitudes (0–5000 m). The results reveal that higher altitudes lead to a notable deterioration in heat exchanger performance. Specifically, compared to sea-level conditions, elevating the altitude to 5000 m results in a concurrent reduction of 23% in the pressure drop and 18% in the heat transfer coefficient. Therefore, while existing fin structures meet low-altitude requirements, they require optimization to adapt to high-altitude environments. However, this optimization process involves evaluating a vast number of design schemes. Traditional computational fluid dynamics (CFD) simulations are often too computationally expensive for this task, creating a significant bottleneck. To address this challenge, we established an efficient optimization framework that integrates numerical simulations, machine learning, and an improved nondominated sorting genetic algorithm II (NSGA-II). Three machine learning models were evaluated, among which the gradient boosting decision tree (GBDT) achieved superior predictive accuracy (determination coefficient R²≈1.0) for both the heat transfer coefficient and pressure drop. Subsequently, multi-objective optimization was realized utilizing GBDT as a surrogate model coupled with the improved NSGA-II. We find that when the pressure drop is comparable to that of the original design, the heat transfer coefficient increases by approximately 23% across all tested altitudes. Conversely, when the heat transfer coefficient remains on par with the original design, the pressure drop decreases by approximately 17%. These findings may help guide the optimal design of next-generation staggered fin heat exchangers suitable for high altitudes.